- 1University of Bergen, Geophysical Institute, Bergen, Norway (noel.keenlyside@uib.no)
- 2Nansen Environmental and Remote Sensing Center, Bergen, Norway
- 3Bjerknes Centre for Climate Research, Bergen, Norway
Climate models suffer from long-standing biases that degrade climate prediction skills. While radically increasing resolution offers promise, we are still many years away from being able to perform operational climate predictions with models that can explicitly resolve the most important physical processes. Here we demonstrate that supermodelling can enhance climate predictions through better using the current generation of models. A supermodel connects different models interactively so that their systematic errors compensate. It differs from the standard non-interactive multi-model ensembles, which combines model outputs a-posteriori. We have developed an ocean-connected Earth System model using NorESM, CESM, and MPIESM in their CMIP5 versions. The model radically improves the simulation of tropical climate, strongly reducing SST and double ITCZ biases. We perform seasonal predictions for the period 1990-2020, initialized through (EnOI) data assimilation of SST. We have performed one forecast per season but are currently extending the ensemble size to ten members. The supermodel shows marked improvement in prediction skill for forecasts started before boreal spring, significantly overcoming the spring predictability barrier. Initial investigation indicates the skill enhancement is connected to better simulation of ocean-atmosphere interaction during the first part of the year, which also leads to improved initial conditions. Our results indicate the importance of better representing the signal-to-noise in the western and central Pacific during boreal spring.
How to cite: Keenlyside, N., Singh, T., Chiu, P.-G., Counillon, F., and Schevenhoven, F.: Overcoming the spring predictability barrier with a supermodel, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12107, https://doi.org/10.5194/egusphere-egu25-12107, 2025.